The Agent Skills Directory

[DATA_EXFILTRATION]: The skill provides detailed code and configuration for accessing highly sensitive personal data via the SensorKit framework. Specifically, it enables collection of keyboard-metrics (which can capture typing patterns and private text input), messages-usage and phone-usage (metadata regarding user communication), and visits (high-resolution location history). Providing an agent with instructions to harvest such data creates a substantial privacy and data exposure risk.
[PROMPT_INJECTION]: The skill establishes a workflow for processing raw data streams from sensors within the agent's context, constituting an indirect prompt injection surface (Category 8).
Ingestion points: SRSensorReaderDelegate methods receiving SRFetchResult in SKILL.md.
Boundary markers: The instructions contain no delimiters or warnings to the agent regarding the untrusted nature of the ingested data.
Capability inventory: The skill leverages the reader's capability to pull high-fidelity user activity data into the agent's environment.
Sanitization: No logic is included to sanitize or validate the content of the sensor samples before they are passed to internal processing functions.

sensorkit